So you click a button, it pops open a text box in a floating window, you type in a question, and the AI replies. This is the most underwhelming implementation of browser-based AI that they could have come up with. Quite literally just gemini.google.com in an iFrame.
The future of web browsing is the tiktok model. Where you don't surf the web, but the web is served to you "algorithmically". Do it long enough, and you'll be serve the pages you want and it will feel like it was your idea all along. Gemini everywhere is the first step.
I stopped using Chrome when they started doing the "logged in to Chrome" thing for all Google services. It seemed likely a creepy step in a vaguely defined, unknown direction. The signal seems stronger now.
This seems ridiculously simple. It doesn’t browse for you in the background or lets you reference tabs etc. This just seems to pass the current page to an llm.
I built an extension like this with Claude-code a few days ago because I wanted to see if I could replace the ai feature of Firefox when I switched to LibreWolf. Turns out, it was quite easy for Claude code.
I want a bit further and tried to get the extension to browse around. Individual actions worked, but I couldn’t get it to follow a plan. In the end I finally looked around the code and Claude had made a huge mess with cursor etc.
The complexity of handling the array of messages was a bit too much for the AI agents.
I now have the same as this Gemini ai though and it CAN click links and it works with ollama too. So more private.
All in a few hours of development.
So I am not impressed by Google here
Similar threads today, sorted by number of comments:
https://news.ycombinator.com/item?id=45292260
https://news.ycombinator.com/item?id=45292163
A danger with google is how flippantly they will ban google accounts for the dumbest things. Now theres a button to livesteam your browsing tied to your google account. I wonder how many people are going to lose 20 years of gmail Gphotos and GDrive files because they accidentally clicked gemini at the wrong moment on the wrong website.
If the forced deprecation of Ublock wasn't enough to get me off Chrome, this sure as hell is.
I was really upset when I found out that my $20 a month Gemini AI Pro subscription only only included privacy features if stopped using the chat history feature.
Gosh I hate google products.
I don’t understand who this is for? I just tried Anthropic’s extension and it feels like writing automated selenium tests.
LLMs interacting with markup is not the best abstraction layer.
> Assouvissez votre créativité sans changer de page
> Have a question about what you're reading? Ask Gemini. It uses the context of your open tabs to provide relevant answers and explanations, keeping you focused.
In France some bits of the page are localized, some are still in English -- doesn't project professionalism or inspire confidence.
For one second, I thought Chrome now supported the Gemini protocol. Then I came back to reality.
> To use Gemini in Chrome on your computer, you need to: Be 18 or over and in the US. Use a Mac or Windows computer. Use the latest version of Chrome. Learn how to update Chrome. Sign in to Chrome. This feature isn’t available in Incognito mode. Learn how to sign in to Chrome. Have Chrome’s language set to English (United States).
Why can't I set Chrome to whichever language I may want and still have that Gemini thing in english?
Yeah, let's have a do-over of this thread. Nice.
https://news.ycombinator.com/item?id=45292260
Maybe someone can post the change log tomorrow and we can do it again.
I'm thinking over the weekend we could post the GitHub merge of these AI features so we can give Google even more exposure.
By Tuesday I hope someone will write a review of these features rehashing the same thing. I'd love to have that be upvoted to the top of HN again.
Microsoft baked in Copilot into Edge more than a year ago.
It was forced into Windows task bar as well.
This seems to be in the same vein.
I hereby declare this to be the future! We made it folks. Time to pack it up. See you in a 2002 LAN party.
> Your web, your control
Typical corporate doublespeak. The web is neither "mine", nor am I ever in control. If anything, the web belongs to corporations like Google. By integrating their text prediction, summarization, and hallucination engine into their web browser, they're further cementing their position of control.
I tried it on this page and says 'I don't have access' [0].
[0] https://drive.google.com/file/d/1Nx4gJA-qWodYWm-SK87Aa63i_jF...
Isn't Google putting AI results at the top some sort of conflict of interest?
Like if users can just get the info they want right at Google.com why would they click through to any of the search results? Isn't that stealing clicks from websites?
I mean, is anyone really surprised that this was going to happen?
Google is about to break even further away in the LLM race with this move, seeing as they will be getting an absolutely, supremely stunning amount of regular and novel data 24/7. Not everyone uses dedicated LLM interfaces, but more people I know use Google search. As Google === Search for so many.
Nevertheless, it is an business savvy move to make, considering the recent ruling by the judge to not force Google to split apart or break up its business w/r/t to Chrome.
How do you turn it off?
Google taking advantage of their anti competitive monopolies
Time for more security researchers to collect more money on data exfiltration reports when attackers instruct and trick LLMs to steal private user information and fall for fake websites generated by AI to accidentally send private information to attackers.
Welcome to the Vibe Browsing security nightmare.
Does anyone know if they are embedding Gemini in chromium and will it eventually be available in nwjs?
> Gemini in Chrome is rolling out to all eligible Mac and Windows users in the US who have their Chrome language set to English. We look forward to bringing this feature to more people and additional languages soon
Knew this was coming thanks to their chrome API's for on-device gen-AI (summarize, translate, generate, etc).
I'm surprised this didn't happen sooner... The amount of data available from Chrome users seems enormous.
I hope Brave deletes this from Chromium if it's present in the source code.
Tangential: Do you have any recommendations for webdev LLM tools?
I would like to inspect some part of the DOM and chat about it with an LLM, including the CSS rules that are applied to each subnode in my selection.
Who wants to bet that Chrome makes this feature impossible to disable?
One thing apple should do is to allow you to chat with the Safari Reader summary to ask questions. The giant machine is too big to move fast.
Given this, i still won't use Chrome.
"This will be so good, just wait! And until then, get used to it!", just like AGI and shitty chatbots that are subtly wrong most of the time.
Obvious mousetrap. Imagine shovelling all this stuff at users for "free". I'm switching to my own Chrominium builds...
Blah. On the one hand, this is where the monopoly power of putting Gemini in Chrome should be looked into by the DOJ. On the other hand, this might make me switch back to chrome.
These are all things Apple could build into safari, but they're nowhere to be seen. They'll be stuck solving yesterday's problems (like building an infinitesimally better camera for the latest iPhone), but not at all integrating any AI into them.
Google had to do this. They cannot die standing watching ChatGPT et. al. eating their ad-free lunch.
Hopefully this will not affect other browser that are downstream of Chromium.
This is why other "AI browsers" that parse and simplify the DOM, then invoke a tool-calling LLM over text are at EOL.
Once Chrome integrates Gemini Live amd treats your browser as a video input stream, it's pixels all the way. No lag, no incorrect clicks on hidden elements.
We need a [US Only] tag on the thread title, I almost got excited
I would love to see usage metrics on that. Probably well below 1% of all browsing sessions, quite possibly even less than 0.1%.
Nobody asked to this. Interpreting websites for its users is categorically not what a web browser is for.
How do we get it out?
Damn, the future is more and more distopic every day.
How is this not stealing clicks from other web pages and advertisers? There is no way that people are forgoing clicking on links at this point if they get the answers right away.
wondering if we can use it with playwright/puppeteer. would be a godsend for scrapers if they can identify useful data.
Given the current err climate of thought purity, doesn't this seem a little too risky of a product to enable?
Ok Google employees, please quit the vote brigading.
Inevitable.
Another reason to ditch Chrome. It's becoming an even more horrible bloatware.
Honestly, it would be great if it were "Gemma in Chrome" instead.
A local model capable enough to do the things that this is designed to do? Yes please.
Gemini in Chrome is a way to increase adoption. Gemma in Chrome is an innovation - a platform that allows developers to build stuff leveraging the local model. A step closer to a world where we can talk to our computers and have them do what we mean instead of what we say.
EU: Open goal and no keeper in sight. Just a small tab. Please.
You didn't want it in your computer, bang, it's there!
You didn't want it in your phone, bang, it's there!
You didn't want it in your browser, bang, it's there!
Next, coming to a fridge near you! /s
Damn, not even a month after getting a butterfly kiss of a slap on the wrist for abusing their monopoly position... and they are already pulling this?
Thank god we have strong regulation in the US to protect us. /s
Google's strategic execution with Gemini over the last two quarters has been impressive. Its deep integration into core products—from the consumer-facing Workspace and Search to developer platforms like Google Cloud and Colab—demonstrates a cohesive, ecosystem-wide approach.
This period marks Google's transition from a preparatory phase to an aggressive market push, which is thus far yielding significant momentum.
This contrasts with the apparent friction in the Microsoft-OpenAI partnership, where long-term strategic alignment seems uncertain. Furthermore, there's a growing perception that competitors like Anthropic are achieving superior performance in specialized domains like software engineering. This suggests OpenAI's current model, which appears heavily focused on optimizing its existing architecture, may be approaching diminishing returns on genuine innovation.
I spent 10 mins trying to find a clear statement of whether Google uses information submitted to Gemini for training and I couldn't find one. It is hard not to come to the conclusion they actively try to obfuscate it because there are many statements that vaguely sound like they should address it but then don't properly do that.
So I would have to suggest, use these features with extreme caution on any page you consider private if you aren't prepared for your private information to get sucked into Google's Gemini training data.